[gmx-developers] Re: New g_wham in CVS (Jochen Hub)

[Michael Shirts]

> Carsten has just uploaded the new g_wham code to the CVS

I'd point out that John Chodera and I have developed a multistate
version of Bennett's acceptance ratio method (MBAR) that will be a
replacement for WHAM in the long run.  The free energy formula looks
exactly like WHAM in the zero-width bin limit -- indeed, WHAM can be
written exactly as a histogram kernel approximation to MBAR.  The
advantages are 1) no histogram bias, 2) no exponential increases in
number of histograms needed with increasing number of states, and 3)
analytical error estimates (no bootstrapping required).  Disadvantages
are 1) iteration over data points rather than histogram bins, which in
some cases can make it take longer -- but we're talking increasing
analysis time from seconds to minutes at worst, so it's not in any way
a bottleneck.

Right now, it's coded up in python, and released at
https://simtk.org/pybmar.   John and I are interested in moving it to
core gromacs tools functionality, but it's slightly lower on the
priority list now (need to get some grad students first!) and will
probably wait until the free energy capabilities of Gromacs are
refactored, which should be soon.  We'd be very happy to help other
people port it, though.  Paper details are at
http:/dx.doi.org/10.1063/1.2978177.

> * proper treatment of autocorrelations (any volunteers? :-) )

Autocorrelations are taken care of in the pymbar scripts.

Best,
Michael

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Michael Shirts
Assistant Professor
Department of Chemical Engineering
University of Virginia
michael.shirts at virginia.edu
(434)-243-1821
Previous Message Next Message 	Back to Sent


> Carsten has just uploaded the new g_wham code to the CVS. It can read
> the gmx4 pull output or the gmx3 pdo file style. In case of gmx4 input
> mode, the umbrella geometry and force constant are taken from the tpr
> files, and displacement from the umbrella center can be provided via the
> pullx or the pullf files.
>
> The new features include:
>
> * error estimation by bootstrapping
> * cyclic (periodic) reaction coordinates
> * tabulated (non-harmonic) umbrella potentials
> * -b & -e, -dt options (to check equilibration)
> * energy output in kJ/mol, kcal/mol, or kT,
> * and g_wham is faster than before.
>
> Wish list:
> * proper treatment of autocorrelations (any volunteers? :-) )
>
> Please let me know if anyone encounters problems or errors. Also
> suggestions for improvements are welcome.
>
> Have fun with it,
> Jochen
>
>
> --
> ************************************************
> Dr. Jochen Hub
> Max Planck Institute for Biophysical Chemistry
> Computational biomolecular dynamics group
> Am Fassberg 11
> D-37077 Goettingen, Germany
> Email: jhub[at]gwdg.de
> Tel.: +49 (0)551 201-2312
> ************************************************
>
>
> ------------------------------
>
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
>
>
> End of gmx-developers Digest, Vol 56, Issue 1
> *********************************************
>